Predicting Appropriate Semantic Web Terms from Words and Table Headers

We are developing an innovative system which takes a set of English words and searches among the ontologies available on the Semantic Web to find the best schemas to encode information associated with the with the words. In other words, a schema written flexibly with ordinary words can be mapped to its canonical form in RDF.

By using the system, the ontology network will gradually evolves, driven by people’s own knowledge and convention, toward a network of concepts resembling those in the real world. People learn relationships among concepts in the world by experiencing (or reading about) how local objects interact. The ontology network will learn by reading RDF documents, which record relationships among concepts in small description experiences.

We will also discuss application of this system that analyzes the words and phrases used to name the columns of a a set of spreadsheets or relational tables and derives a consistent schema for the information.